Creator profiles now show real avatars, image and video archives, search, sorting and work previews.
Aligned navigation, typography, community and information pages across desktop, mobile and four languages.
New frontend and automatic video publishing
New layouts for discovery, community, characters, creators and information pages, with mobile and dark-mode support.
Creator registration, RSSHub discovery, AI review and automatic publishing are connected, with release notifications sent to the Feishu group.
Creator registration, video sync and automatic publishing
Join the sync queue with a Bilibili UID or profile URL and a short introduction.
RSSHub discovers recent uploads. Each video is checked for uploader identity, cover art and preview frames; only AI girl fan works are included.
Approved videos and creator profiles enter the publishing queue automatically. Review and release results are sent to the Feishu group.
New Frontend and Theme Browsing
Launched the Discover homepage, character profiles, theme collections, and full artwork detail pages, while keeping existing artwork URLs and comment records.
The gallery now uses a single “See more” flow for continuous browsing; the filter bar, dropdowns and preview lightbox share one set of components.
Keeps the grid-paper and hand-drawn paper-card look, with support for phones, dark mode and four interface languages.
On-site submissionsAdded bot verification and rate limiting
The quick on-site submission form now uses Cloudflare Turnstile bot verification — it usually passes on its own, and a single click is all it takes when it doesn't.
Each IP can submit up to 10 times per hour; go over that and you'll be asked to try again later.
The verification widget stays light at all times, matching the site's color scheme.
QQ group submissionsOpened
In the submission group, @ the bot and send it the submission command followed by the title and character, with the image attached in the same message; if any parameter is off, it flags each one.
Character names accept the usual aliases and the -chan suffix as well, e.g. DeepSeek Chan.
Group submissions allow 10 images per clock hour in total; the count resets on the hour, and the reply shows how many are left for that hour.
Submissions from all three channels — on-site, GitHub and QQ — go into one review queue, then publish in batches once approved.
Front endMoved to Astro
Every page is now built statically with Astro, with page structure and URLs unchanged.
The old page-generation scripts have been retired.
Topic pagesLaunched, with a new topic carousel on the home page
Added topic pages that group related works by theme so you can browse them together.
A topic carousel on the home page puts the newest topic right up front.
English / JapaneseInterface language support goes live
Added English and Japanese interfaces: the three-language sticker on the right of the top bar switches with a single click, your choice is remembered on this device, and the ?lang= parameter and your browser language are respected too.
Interface copy for all seven handwritten pages and 250 detail pages now runs through the translation layer; category names, category terms, and source and license labels are translated too.
Work titles, tags and detail-page body text are content data, so they stay in the original Chinese; under Japanese the site automatically switches to a Japanese font stack.
Submission formSimplified
GitHub submissions now just need the image uploaded plus an image name and character; the one-line description is optional.
Tags, category, source, license and detail-page body text are now filled in by the maintainer while reviewing the image, so submitters no longer need to understand the whole listing spec up front.
The on-site quick submission and Feishu channel still aren't open — the real way in is still the GitHub Issue form.
Custom 404The page goes live
Added the site's own 404 page: a big 404, a few possible causes, and three ways out — back to the home page, browse the categories, or go submit something.
The page still returns a 404 status code and is marked noindex, so search engines won't treat it as a normal page and index it.
Like every other static page, the 404 page goes through gzip pre-compression and build validation; if it's missing from the output, both the build and the release fail outright.
Image-by-image visual reviewReclassifying, and a new Comics category
Went back through the images for all 191 works and re-sorted them by what's actually in the picture: Reaction images / Fan art / Character designs / Comics.
Created a new Comics category for multi-panel sequences, picture stories and webtoon-style strips.
Detail page copy is now a first-person comment from DeepSeek Chan, handwritten for each image.
The first 59 images — which had no titles or tags and had never gone live — got their titles and tags filled in from the pictures this time and were all brought in, taking the total from 191 to 250.
The originals for these 59 images are now self-hosted on this site, so downloads no longer depend on the upstream archive.
Fixed the missing space in titles and image alt text where tags were glued to "AI Chan fan art" (e.g. glmAI Chan fan art → glm AI Chan fan art).
Architecture boundariesGround rules and data boundaries land
Added a spec document for "Architecture boundaries".
Moved characters and categories out of the code into JSON lists.
Added reference-integrity assertions to the build.
Third batchSubmissions added(#19–#22)
Added 6 images.
Introduced two new characters: StepFun Chan and GLM Chan.
First editionLoading optimizationsShipped
Added thumbnail and detail-image derivatives for submissions.
Release bundle: 73.7 MB → 8.3 MB.
This page is an excerpt; the complete changelog is archived at archive/2026-09-24/CHANGELOG.md in the repo.